Back to blog
Tech & AI

Google Tests Turning Off Gemini Image Watermarks — SynthID Likely Stays

Google appears to be testing a "watermark setting" that would let users turn off the logo in the bottom right corner of images generated by Gemini, after the option was spotted in app code (version 17.43.11).

Only the visible watermark is likely to go

Josh Woodward, who leads Gemini and AI Studio, said users had frequently requested removal of the visual watermark.

Given that, the update is likely to remove only the visible watermark while retaining two things:

  • The invisible watermark SynthID
  • Metadata-based C2PA provenance information

That distinction is the substance of the story. A visible logo and machine-readable provenance serve different purposes.

  • The visible logo tells a viewer the image is AI-generated — but damages design and constrains commercial use
  • SynthID and C2PA let machines verify content provenance — with no effect on how the image looks

The change reads as a compromise: accommodating the usability request while preserving traceability.

However, neither the release timing nor whether it applies to previously generated images has been confirmed.

What marketers should note

Removing the watermark does not remove disclosure obligations. Even with the visible logo gone, regulatory requirements to disclose AI-generated content in advertising remain. How those are being operationalised is covered in Google Rolls Out AI Content Labels Across Its Advertising Platforms.

Build a labelling system for bulk operations. Watermarking individual images and declaring AI use at campaign level are different tasks. For bulk AI labelling in file-based workflows, see DV360 Structured Data Files v10.1.

Manage assets on the assumption SynthID persists. If the invisible watermark is retained, AI generation can still be verified later. Operating on the premise that "no logo means nobody will know" is a risk position.

Remember nothing is confirmed. This is a feature found in app code, with release timing and retroactive application both undecided. It is too early to design a workflow that assumes it.

Frequently Asked Questions

How was the Gemini watermark setting discovered?

It was spotted in app code version 17.43.11 as a "watermark setting" allowing users to turn off the logo in the bottom right corner of generated images.

Will all watermarks be removed?

Likely only the visible one. The invisible SynthID watermark and metadata-based C2PA provenance information are expected to remain. Gemini lead Josh Woodward noted users frequently requested visual watermark removal.

Does it apply to previously generated images?

Not confirmed. Both the release timing and whether the setting applies retroactively to existing images remain undecided.

Does removing the watermark remove disclosure duties?

No. Visible logo removal is separate from regulatory disclosure obligations, and if invisible SynthID persists, AI generation can still be verified afterwards.

Where does your own site stand?

To apply what you just read to your own site, start with a free audit of where things are now.

A strategist replies within 24 hours on business days.

Read next